Title: Building AI in India, for India and the World: Challenges, Opportunities, and the Road Ahead

Speaker: Niket Agarwal, Senior Distinguished Engineer, NVIDIA.

Abstract: India is roughly 18% of the world's population and under 1% of the data that large language models are trained on -- an asymmetry usually framed as a data problem, but which is equally a systems one: Indic scripts cost four to thirteen times more tokens than English for the same content, so usable context collapses, KV-cache and prefill costs rise in proportion, and efficiency techniques validated on English (quantization, speculative decoding, distillation) degrade in ways that automatic metrics understate by close to an order of magnitude.

The opportunity is that the frontier has moved: capability now comes from post-training (reinforcement learning, verifiable environments, reward design), rather than from parameter count, and that frontier is reachable from India in a way trillion-parameter pretraining is not, provided we build RL compute in country, open environments that encode Indian tasks, and evaluation we can trust -- the binding constraint, since you cannot optimize against a reward you cannot measure.
These constraints are not parochial: severe cost sensitivity, latency budgets under real network conditions, code-mixed and romanised input, voice as the primary interface, and domestic control that mandates in-country inference are the conditions the world's next several billion users will impose, so building for them here produces systems research that travels and the decisions taken in the next twelve months will determine what India runs for the rest of the decade.

Speaker Bio: Niket Agarwal is a Senior Distinguished Engineer at NVIDIA, where he leads initiatives at the intersection of AI infrastructure, large-scale model development, and accelerated computing. He is driving NVIDIA's India ecosystem engineering strategy, working closely with startups, enterprises, academia, and government organizations to accelerate AI adoption and innovation across the region. Niket also leads NVIDIA's India-focused Nemotron research initiatives, driving collaborations on multilingual AI, Indic language models, data ecosystems, and sovereign AI efforts. His work spans large language models, agentic AI systems, inference and training optimization, and the deployment of AI infrastructure at scale.

Prior to NVIDIA, Niket spent nearly a decade at Google building large-scale distributed systems and infrastructure, followed by engineering leadership roles at Meta. He holds a PhD in Computer Engineering from Princeton University and a bachelor's degree from IIT Kharagpur, with specialization in systems and computer architecture.

Niket is particularly passionate about advancing AI for India and enabling the next generation of AI platforms, applications, and talent.

Title: Limitations of Quantum Advantage and Implications

Speaker: Apoorva D. Patel, Emeritus Professor, Centre for High Energy Physics, Indian Institute of Science, Bangalore.

Abstract: Quantum computation is universal, but quantum advantage is not. Applications of quantum technology will be practical only in situations where the advantage offered by them is large enough to offset their technological cost. Genuine quantum features appear in algorithms and devices only when they involve the fundamental Planck constant in some form or the other. Explicit scenarios illustrating this constraint will be described. The implication is that quantum technology will be used only as special purpose subroutines in hybrid classical-quantum settings.

Speaker Bio: Apoorva D. Patel is an Emeritus Professor at the Centre for High Energy Physics at the Indian Institute of Science, Bangalore, and is involved with the IISc Quantum Technology Initiative. He is also a Visiting Professor at the ICTS-TIFR, Bangalore. He obtained his MSc in Physics (1980) from the Indian Institute of Technology Bombay, and his PhD in Physics (1984) from the California Institute of Technology, USA. He has extensively worked on lattice gauge theory investigations of Quantum Chromodynamics. He is also notable for his work on quantum algorithms, and the application of information theory concepts to understand the structure of genetic languages. He was exposed to the subject of quantum computation during his graduate student days at Caltech, and he has been teaching the subject at IISc for more than 25 years.